diff options
| author | Nadav Rotem <nrotem@apple.com> | 2012-08-18 02:43:28 +0000 |
|---|---|---|
| committer | Nadav Rotem <nrotem@apple.com> | 2012-08-18 02:43:28 +0000 |
| commit | b5838689c6e204af3677df0ae10167f6070aba00 (patch) | |
| tree | a29c2cc54eadc0f0527df47a04fa9b5d489e35ac /test | |
| parent | fd45fa1503de725801be3db33c7e860298fc82a3 (diff) | |
| download | external_llvm-b5838689c6e204af3677df0ae10167f6070aba00.zip external_llvm-b5838689c6e204af3677df0ae10167f6070aba00.tar.gz external_llvm-b5838689c6e204af3677df0ae10167f6070aba00.tar.bz2 | |
The X86 backend has a number of optimizations for SETCC nodes which use
arithmetic instructions. However, when small data types are used, a truncate
node appears between the SETCC node and the arithmetic operation. This patch
adds support for this pattern.
Before:
xorl %esi, %edi
testb %dil, %dil
setne %al
ret
After:
xorb %dil, %sil
setne %al
ret
rdar://12081007
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@162160 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test')
| -rw-r--r-- | test/CodeGen/X86/2012-08-16-setcc.ll | 45 | ||||
| -rw-r--r-- | test/CodeGen/X86/fold-load.ll | 4 |
2 files changed, 47 insertions, 2 deletions
diff --git a/test/CodeGen/X86/2012-08-16-setcc.ll b/test/CodeGen/X86/2012-08-16-setcc.ll new file mode 100644 index 0000000..ed51156 --- /dev/null +++ b/test/CodeGen/X86/2012-08-16-setcc.ll @@ -0,0 +1,45 @@ +; RUN: llc < %s -mtriple=x86_64-apple-macosx | FileCheck %s + +; rdar://12081007 + +; CHECK: and_1: +; CHECK: andb +; CHECK-NEXT: cmovnel +; CHECK: ret +define i32 @and_1(i8 zeroext %a, i8 zeroext %b, i32 %x) { + %1 = and i8 %b, %a + %2 = icmp ne i8 %1, 0 + %3 = select i1 %2, i32 %x, i32 0 + ret i32 %3 +} + +; CHECK: and_2: +; CHECK: andb +; CHECK-NEXT: setne +; CHECK: ret +define zeroext i1 @and_2(i8 zeroext %a, i8 zeroext %b) { + %1 = and i8 %b, %a + %2 = icmp ne i8 %1, 0 + ret i1 %2 +} + +; CHECK: xor_1: +; CHECK: xorb +; CHECK-NEXT: cmovnel +; CHECK: ret +define i32 @xor_1(i8 zeroext %a, i8 zeroext %b, i32 %x) { + %1 = xor i8 %b, %a + %2 = icmp ne i8 %1, 0 + %3 = select i1 %2, i32 %x, i32 0 + ret i32 %3 +} + +; CHECK: xor_2: +; CHECK: xorb +; CHECK-NEXT: setne +; CHECK: ret +define zeroext i1 @xor_2(i8 zeroext %a, i8 zeroext %b) { + %1 = xor i8 %b, %a + %2 = icmp ne i8 %1, 0 + ret i1 %2 +} diff --git a/test/CodeGen/X86/fold-load.ll b/test/CodeGen/X86/fold-load.ll index c961f75..d836665 100644 --- a/test/CodeGen/X86/fold-load.ll +++ b/test/CodeGen/X86/fold-load.ll @@ -57,13 +57,13 @@ entry: %0 = load i32* %P, align 4 %1 = load i32* %Q, align 4 %2 = xor i32 %0, %1 - %3 = and i32 %2, 65535 + %3 = and i32 %2, 89947 %4 = icmp eq i32 %3, 0 br i1 %4, label %exit, label %land.end exit: %shr.i.i19 = xor i32 %1, %0 - %5 = and i32 %shr.i.i19, 2147418112 + %5 = and i32 %shr.i.i19, 3456789123 %6 = icmp eq i32 %5, 0 br label %land.end |